The Impact of the Minimum Wage on the Destruction and Creation of Products *

Transcription

1 The Impact of the Minimum Wage on the Destruction and Creation of Products * Roberto Alvarez University of Chile Lucas Navarro Universidad Alberto Hurtado Abstract We study the impact of changes in the legal minimum wage on the creation and destruction of products at the firm level. This can be a relevant way for increasing firm productivity and for explaining why the raises in the minimum wage may have minor effects on employment. Our identification strategy exploits as a quasi-experiment a large and 3-year predetermined increase in minimum wages during in Chile and the differences in products exposure to these changes. Our main results indicate that increases in the minimum wage raise the destruction of products that are more unskilled labor intensive and that augments the introduction of products that are more intensive in skilled workers. The impact is economically relevant. An annual nominal increase of about 10% in the minimum wage as occurred in this period increases the probability of dropping unskilled labor products in 5 percentage points and reduce the probability of creating unskilled labor products in 3.3 percentage points. Our results are robust to sample selection issues, to the introduction of controls for potential confounding factors and to the consideration of alternative definitions of exposure. * We thank the financial support from CAF and the valuable comments and suggestions received from participants at conferences in the IADB, University of Chile and CAF, specially Eugenia Andreasen, Marcela Eslava, Oscar Mitnik and Mariano Tappata. 1

2 1. Introduction There is an abundant and controversial literature about the impact of legal minimum wages on labor markets; in particular, several studies have analyzed the effects on wages, employment, young workers labor outcomes, and on poverty and inequality. Recent literature reviews indicate that changes in minimum wages are associated with minor reductions in employment (Belman and Wolfson, 2014; Neumark, et al. 2014). In some cases, this effect has been found to be positive (Card and Krueger, 1995; Machin and Manning, 1996). The main explanation for these findings is that in monopsonistic markets, a rise in minimum wage reduces market power and firms expand employment (Schmitt, 2013) 1. Surprisingly, there is not much evidence on the impact of the minimum wage using firm-level data. In recent papers, Mayneris et al. (2014) study how changes in minimum wages in China affect firm s productivity and survival, and Draca et al. (2011) look at the impact of the minimum wage on wages and profitability for U.K. firms. The study of firm level responses to changes in labor costs would contribute to a better understanding of the mechanisms behind the aggregate relationship between the minimum wage and employment. Microeconomic studies can also help to illustrate the heterogeneous impact of labor market policies and would eventually be useful for implementing complementary policies. 1 See De Fraja (1999) for a model where this result is not due to monopsony. 2

3 In the case of the minimum wage, it has been argued that small effects on employment, survival and profitability may be due to adjustment by firms through different mechanisms (Schmitt, 2013; Hirsch et al. 2015). One of them is increasing their productivity when firms face this negative shock 2. There are several sources of productivity growth: investment in new technologies, probably more capitalintensive, doing a better and more careful selection of workers (Autor et al., 2007), or changing the product mix by concentrating their resources in the more productive products (Bernard et al., 2010). Based on recent theoretical and empirical literature on multi-product firms, in this paper we analyze how changes in the minimum wage are associated with changes in the mix of products, i.e., the introduction of new products and the dropping of some of them, as a potential mechanism for avoiding the negative effects of labor market regulations. In related theoretical models (Bernard et al. 2010), firms endogenously sort across products and variations in the product mix may have important positive effects on firm and aggregate productivity. The creation and destruction of products may occur as a result of policy changes, such as trade liberalization (Bernard, et al. 2011) or the exposure to competitive pressures (Mayer, et al. 2014). In both cases, and similar to what we expect in the case of a negative shock as the increase in the minimum wage, firms react by increasing production of 2 Other alternative, in less competitive industries, is that firms raise prices and trespass to consumers the increase in labor costs Aaronson (2001); Wadsworth (2010) 3

4 their most productive, and hence higher-profits goods, that would allow them to survive in the new environment. We contribute to three strands of the literature. First, we analyze the heterogeneous effects of labor market regulations on productivity. Although Alvarez and Furentes (2017) find a negative effect of increases in the minimum wage on Total Factor Productivity (TFP) in a similar period to that analyzed here, we investigate how firms adjust internal resources to reduce the negative impact of this labor regulation. Second, in terms of impact evaluation of labor market policies, we provide novel firm-level evidence on the impact of minimum wages in a new dimension: creation and destruction of products. To our best knowledge, this is the first paper looking at this relationship. Second, regarding the literature of multi-product firms, we analyze an under-studied potential determinant of products creation and destruction. Most of the previous literature has focused on the effects of trade liberalization (Nocke and Yeaple, 2014; Goldberg et al., 2010; Mayer et al., 2014) and it shows, with the exception of Qiu and Zhou (2013), that changes in the product mix may have large and positive effects on productivity. We take advantage of a large increase of the minimum wage in Chile taking place between 1998 and Interestingly, by the end of 1997 the Minister of Finance announced a predetermined increase in the minimum wage for this three-year period. This was in sharp contrast to how the policy was conducted in previous and 4

5 subsequent periods where the Government announced and implemented the minimum wage on a yearly basis. This change is important for our identification strategy because current economic conditions, such as the evolution of TFP and other macro variables, should relate more closely to the annual, rather than to the triennial, setting of the minimum wage. As shown in Figure 1, the annual real increase in the minimum wage was, on average, 7.3% in (approximately 10% in nominal terms), well above the increase in preceding and succeeding years. During the beginning of the 1990s, the minimum wage increased, on average, 4.2% per year and 2.5% per year only during the period Our identification strategy exploits as a quasi-experiment this large increase in minimum wages and the differences in products exposure to these changes. We consider as exposed products to those that are more intensive in unskilled labor (production or blue-collar workers) because these unskilled-intensive products should experience a large increases in production costs due to the increase in minimum wage. In extreme, if some products were produced only with skilled workers receiving wages well above the minimum wage, these would not be affected by this regulation. As we do not observe product-specific input intensities, we take advantage of the information for single-product firms and we use them as a proxy for unskilled labor intensity of products produced by multi-product firms. The 3 See Alvarez and Fuentes (2017) for a discussion on this issue and how the increase in minimum wage was well above the labor productivity growth. 5

6 underlying assumption of this procedure is that for each produced product, multiproduct firms use the same technology as single-product firms. In different frameworks, Ma, et al. (2014) and De Loecker, et al. (2016) use the same assumption. Our main results indicate that increases in the minimum wages effectively increase the destruction of products that are more unskilled labor intensive and that firms introduce products that are more intensive in skilled workers. The impact is relevant. An increase of about 10% as that occurred between 1998 and 2000 increase the probability of dropping unskilled labor products in 5 percentage points and reduce the probability of adding unskilled labor products in 3.3 percentage points. Our results are robust to several robustness checks such as, sample selection issues, controlling for confounding factors and to alternative definitions of exposure. We also show evidence that our findings do not seem to be explained by differences in previous trends for exposed and non-exposed firms. The paper is structured as follows. In the second section, we present the data. In the third section, we discuss the methodology. In the fourth section, we present our basic results and the robustness checks. The fifth section concludes. 2. Data We use data from the national annual manufacturing survey (Encuesta Nacional Industrial Anual, ENIA), managed by the official Chilean statistics agency (INE). The 6

7 unit of observation is a plant with ten or more employees and there are on average more than 4,000 plants per year in the sample. The ENIA, in addition to information on plant characteristics, provides data about plants products. The latter information is contained in Formulario Número 3 of the survey (Form 3, from now on, F3) and allows us to identify the specific goods that the plants produce 4. The information on plant products is available up to 2003 but there was a change in products classification in Even though there are harmonization tables for the two product classifications, the prevalence of product mix changes that would result for 2001 seems too high to be reliable. For this reason, we consider plants that introduced product mix changes only during the period Products are defined according to a local classifications denoted by CUP (Unique Products Classification). The product information is more disaggregated than a seven-digit Second Revision International Standard Industry Classification (ISIC). Hereafter, we will refer to the more disaggregated definition of a product as "product" or "ENIA product." It is possible to assign the products to different sevendigit and more aggregate ISIC categories. We will refer to two-digit ISIC categories as "sectors" and four-digit ISIC categories as "industries." There are 10 sectors, 91 4 It should be noted that more than 95% of the firms produced in a single plant in 1996, the only year with firm and plant level information available. For this reason, we will use the terms firm and plant interchangeably. 7

8 industries, 257 five-digit ISIC categories, 587 six-digit ISIC categories and 2112 ENIA products in the pooled sample. In Table 1, we present information on the number of plants and products per year under alternative product aggregations for the sample used in this study. The total number of firms decreased significantly over the years, by much more than the total number of products, which remained approximately constant considering the two extreme years in our sample. Consequently, the average number of products per firm decreased from 2.6 to 2.1 between 1996 and The data on plants products by year allows us to identify the creation and destruction of a product over time. Thus, our definition of product creation (adding) considers the case of firms producing a product in year t, which was not produced in t-1. Similarly, product destruction (dropping) refers to a product that was produced in t-1 but not in t. In Table 2, we present information on the percentage of products added and dropped between two consecutive years in relation to the total number or plant-products observations in each year. On average, 21% add 15% of the plant-products in the sample were dropped and added every year, respectively. Approximately more than half of the products drop rate and two fifths of the entry rate come from the exit and entry of firms to the sample, correspondingly. Interestingly, the add and drop rates reach their lowest and greatest levels, respectively, in 1998, one year after the new three-year minimum wage policy was 8

9 announced. In Table 3, we show the incidence and relevance of product mix changes across sectors. The figures show that the incidence of product mix changes is higher in the Wood and Metallic sectors and lower in Food and Beverages. 3. Methodology We are mainly interested in analyzing how variations in minimum wage affects product switching. To do that, we estimate the following equations: =1 = + + ( ) ( ) + + =1 = + + ( ) ( ) + + Where αpf is a set of firm-product fixed effects, αjt is a set of industry-year fixed effects for capturing shocks that are common to products belonging to the same industry j, MW is the minimum wage (in logs), Exp is our measure of exposure, and X is a vector of firms characteristics, including the measure of exposure. As it can be appreciated in the equations, we allow to the minimum wage affect with a one year lag to entry/exit products decisions. The exposure variable is measured before the change in the minimum wage, specifically in Following Bernard et al (2006), who analyze the exposure of manufacturing plants in the U.S. to the imports competition from low-wage countries, our measure of exposure is given by: = [ / h ] 9

10 This measure, the wage-bill ratio between unskilled and skilled workers, captures both differences in wages and unskilled workers intensity at the firm-level. Given that for multi-product firms, we only observe the wage-bill ratio at the firm-level and not at the product level, as desired, we use the product information for this ratio for single-product firms. Under the assumption that unskilled intensity for a determined product is similar between single-product and multi-product firms 5, we can define our measure of exposure as the average for single-product firms that produce the same product that multi-product firms. The rationality behind this assumption can be explained considering the case of a firm producing two products, denoted by 1 and 2. The wage-bill ratio of the firm (ωf), will be the weighted average of the wage-bill of the workers utilized in the production of both goods. This is: = + When α1 1, then ω ω1. Then if the multi-product firm produces product p, then we use the average of the exposure variable across all of single-product firms that produce p. To be consistent, we compute this variable also for the first year of our sample period. In the case that other single-product firms did not produce some products, we use the information for single-product firms producing those products defined at higher 5 Ma et al. (2014) use a similar assumption for unobserved capital intensity of new exporters in China. De Loecker et al. (2016) rely on the same assumption to estimate product-level production functions. 10

11 levels of aggregation, i.e. at the 6-digit and 5-digit ISIC levels. Overall, we are able to identify the product exposure variable for more than 93% of the observations in our sample. Figure 2 presents evidence of how good is our prediction of the wage-bill ratio for multi-product firms. The solid line shows the Kernel density of the observed exposure variable across plants in The dashed line shows the Kernel density of the corresponding predicted exposure computed as a weighted average of the product level exposure values obtained using our proposed measure.. It can be observed that the methodology predicts reasonably well the observed distribution of the plant level exposure variable. Given that increases in wages should affect more to unskilled-intensive products, ewe expect that an increase in the minimum wage reduces the probability of product entry (δ1<0) and that increases the probability of product exit (δ2>0). The interaction of minimum wage with exposure implies that the effect will be higher for unskilledintensive products. In fact, the marginal change of an increase in the minimum wage will be given by: h = ( ) ( ) = The model is estimated using a linear probability model, and not a Probit or Logit, because the linear model allows to introduce fixed-effects to control for unobserved heterogeneity. In some of our regressions, we also introduce categorical variables according to the distribution of. In particular, we use a dummy for 11

12 products with unskilled intensity in the superior third of the distribution. Also, we check the robustness of our results to changes in how exposure is measured and by doing some placebo tests. 4. Results Table 4 presents our basic results for product destruction and creation. We present results using the continuous measure of exposure as defined in the previous section with and without control variables. We also use a discrete measure of exposure Q3EXP, which is a dummy equal to 1 if the product is in the upper tercile of the distribution of the exposure variable and 0 otherwise. Our findings for product destruction suggest that an increase in minimum wage raises the probability of dropping products, and the impact is higher for more unskilled labor-intensive products. In the case of product creation the parameter for the interaction between minimum wage and exposure is negative and significant, indicating that an increase in labor costs reduce the probability of introducing a product that is more intensive in unskilled labor. In term of the control variables, their introduction does not affect the sign and significance of our variables of interest. These results show that larger and more productive firms are less likely to drop products. The quantitative impact is relevant, but not dramatic. Evaluating the results with the dummy for exposure, an increase of 10% in nominal minimum wage, raises the product destruction of unskilled labor-intensive products in about 5 percentage 12

13 points, compared with a sample average for plants in the third quartile of the exposure variable of 22%. In the case of product creation, the impact is slightly lower about 3.3 percentage points and the sample average is 15%. We undertake several robustness checks of our results. First, we only run the model for surviving firms because product destruction and creation can be driven by exit and entry of the firms. The results, presented in Table 5, are very similar to those of Table 4 and indicate that a 10% increase in minimum wage raise in 3.1 percentage points the probability of dropping unskilled labor-intensive products and a 2.7 percentage points decrease in the probability of adding unskilled labor-intensive products. This, compared with the average 9.5% and 9.7% product drop and add rates among survivors in the treated group in the data, suggests the existence of relatively larger effects of the minimum wage among survivors than in the whole sample. Second, we check that our results are not driven by a differential behavior of single versus multiple products plants. Table 6 and Table 7 reports the estimation results of our base model for the sample of single and multi-product plants, respectively. The results are qualitatively the same as those of Table 4, but with reduced statistical significance (especially for single-product plants) perhaps due to the reduced samples size. 13

14 Third, we also estimate the model to analyze the effect of the minimum wage on changes in the exported product mix (see Table 8). Typically, firms export their most productive products, which are less likely to be affected by changes in the minimum wage. Also, the existence of fixed costs of exporting, would suggest more persistence of selling products in international markets. The evidence suggests some support for this idea. We find that exported products are not affected by changes in minimum wage. We also investigate if products that were exported in 1996, were more or less likely to be dropped with the large minimum wage increases and if this effect depends on the exposure variable also. We estimate a model with a triple interaction (our exposure variable, the minimum wage and a dummy Exported Product indicating if the product was exported in 1996). Table 9 presents the results. First, the coefficient of the interaction term Exported Product and MW shows that exported products in 1996 were less likely to be dropped later one. Second, as expected for unskilled intensive products, the coefficient of triple interaction term suggests that this effect was attenuated in more exposed products. Summarizing, the results of Table 8 and Table 9 indicate that exported products were less affected by the minimum wage increases. Fourth, given that this period coincides with the Asian crisis that affected Chile, and more unskilled intensive products may be more affected by the contraction in 14

15 the economic activity, we introduce an interaction term between sectorial GDP growth and our measure of exposure. Table 10 shows the results. Previous evidence by Bernard and Okubo (2016) find that recessions are periods of increased product creation and destruction in Japan. In our case, the negative coefficients for the interaction terms of our exposure variable with the growth of GDP (dgdp Sector) are therefore consistent with their evidence. More interestingly, including this interaction term does not affect our main results related to the effect of variations in the minimum wages on products creation and destruction. We performed a similar exercise using the interaction of our exposure variable with aggregate GDP growth and the results are basically unchanged. Fifth, given that firms may change products according to their product scope, we also introduce an interaction term of the minimum wage with a measure of distance to the product scope. We define product scope as the plant s average 7-digit ISIC codes. Distance to scope is the log difference between the product code and the plant s product scope. The larger this variable the more different the product from the plant s product mix. The intuition is that variations in the last digits of the firms product code imply lower dispersion in a product scope than variations in the 6 or 5-digit level of their ISIC codes. We would expect that increases in the minimum wage would lead to increased destruction of products away from the scope and reduced creation of those products. 15

16 In Table 11 we show our findings in the case of including the interaction between the minimum wage and distance to scope. An alternative explanation for our previous findings is that an increase in labor costs may induce firms to rationalize the product mix and concentrate resources on the products closer to their scope or competence as modelled by Eckel and Neary (2010). We find, as expected in the case that firms specialize in core competences, that the parameter for the interaction term is negative for product creation and positive for product destruction. More importantly, our results for the interaction effects of the minimum wage and unskilled labor-intensity holds. These last results do not necessarily imply that the large increase in the minimum wage only affected the creation and destruction of products away from the core competencies of the Chilean firms. On the contrary, if we restrict the attention to the effect of the minimum wage on core products mix changes we also find a significant impact on those more unskilled intensive. We define core products to those representing at least 75% of the firm sales. The results are shown in Table 12 and suggest that the minimum wage led also to core products mix changes. We present in Table 13 a placebo test to check that our results are not driven by any potential spurious relationship between the minimum wage and changes in the product mix. To do that, we generate random allocations of products unskilledintensity (exposure) for multi-product firms. Our findings indicate that the interaction 16

17 between minimum wages and these randomly assigned intensities is not significant. Then, changes in minimum wages seems to be effectively associated with variation in product mix that are dependent on the unskilled workers intensity of the products. In Table 14, we consider instead as exposure variable, the bite of the minimum wage, defined as the fraction of workers earning one and up to 1.2 minimum wages in 1996 at 3-digit ISIC level industries, using individual data from the 1996 National Survey of Social-Economic Characterization (CASEN). We use the continuous value of this fraction. The results are presented in Table 14 for both for the total sample (columns 1 to 4) and the sample of surviving firms between two consecutive years (columns 5 to 8). The results are unchanged and confirm the findings that an increase in minimum wage is associated with dropping unskilled-intensive products and adding more skilled-intensive products. Our final experiment consists of estimating our model using a more restrictive definition for product mix changes, that is considering only product additions and dropping at the more aggregate 6-digit ISIC level. That is, it could be the case that a firm creates a product at the 7-digit level but not at the 6-digit level, if the created product has the same 6-digit code than the rest of the prevailing products of the firm. Considering too disaggregate definitions of product may also be related to spurious product innovations that the 6-digit product definitions help to mitigate. 17

18 Table 15 presents our findings using this alternative aggregation for defining products. The results, again, are mostly unchanged. Finally, we show some evidence regarding the parallel trends assumption between more exposed and less exposed firms. One traditional concern with this differences-in-differences approach is that the impact could reflect uncontrolled previous trends in the interest variable for treated and control groups. We do not have enough data on product creation and destruction to look at the period before the change in minimum wage, but we can analyze the behavior of related variables. Consistent with our estimations, we define treated firms are those with unskilled intensity in the superior third of the distribution and as control the rest of the firms. First, we do not see differences in evolution of the average number of skilled and unskilled workers as it would be if for any other reasons both groups of firms were changing products and unskilled-intensity before the change in minimum wage (Figure 3). Also, other variables that may be related to changes in the product mix, such as exporting and technology investment (measured as purchases of foreign licenses), do not show differences in the previous period to changes in the minimum wage (Figure 4) 6. In sum, our general evidence seems to be robust and consistent with the idea that lower employment effects at the firm-level due to increasing minimum wage 6 We show the percentage of exporters and firms purchasing licenses in both groups. Similar evidence is found for the exports to sales and licenses to sales ratios. 18

19 may be explained by firms adjustments in the product mix aimed to increase productivity. In fact, as it is shown in Figure 5, during this period, firms that introduced changes in the product mix outperformed in terms of productivity to those that did not introduce variation in their products. 5. Conclusions There is a large debate on the effects of minimum wages on employment, but few empirical evidence about how firms respond and adjust to these shocks. We contribute to this literature by studying the impact of changes in minimum wage on the product creation and destruction at the firm level. Our identification strategy exploits as a quasi-experiment a large and 3-year predetermined increase in minimum wages during and the differences in products exposure to these changes. Our main results indicate that an increase in the minimum wages effectively increase the destruction of products that are more unskilled labor intensive and that firms introduce products that are more intensive in skilled workers. The impact is economically relevant, but not dramatic. In our basic regressions, we find that an annual nominal increase of 10% as that occurred between 1998 and 2000 raises the probability of dropping products of low skill intensity in 5 percentage points and decrease the probability of adding those sort of products in 3.3 percentage points. 19

20 Our results are robust to sample selection issues, controlling for confounding factors and to alternative definitions of exposure. Then our evidence is consistent with the idea that lower employment effects of higher minimum wage at the firm-level may be explained by adjustment mechanisms for increasing productivity. In our case, previous evidence has shown that changes in the product mix may be an important way to increase productivity, by reallocating resources within the firm. Obviously, there are other adjustment mechanisms that may be explored. However, it seems that product creation and destruction is a relevant and robust one. 20

21 References Aaronson, D. (2001). Price pass-through and the minimum wage. Review of Economics and statistics, 83(1), Álvarez, R., & Fuentes, R. (2017). Labor market regulations and productivity: Evidence from Chilean manufacturing plants. Forthcoming, Economic Development and Cultural Change. Autor, D. H., W. Kerr and A. D. Kluger (2007). Does Employment Protection Reduce Productivity? Evidence from US States. The Economic Journal 117(521): Belman, D., & Wolfson, P. J. (2014). What does the minimum wage do?. WE Upjohn Institute. Kalamazoo, Michigan. Bernard, A. B., Jensen, J. B., & Schott, P. K. (2006). Survival of the best fit: Exposure to low-wage countries and the (uneven) growth of US manufacturing plants. Journal of international Economics, 68(1), Bernard, A., Redding, S., and P. K. Schott Multiple-Product Firms and Product Switching. American Economic Review 100(1): Multiproduct Firms and Trade Liberalization. Quarterly Journal of Economics 126(3): Bernard, A. and T. Okubo (2016) Product Switching and the Business Cycle, NBER Working Paper

22 Beyer, H. (2008). Mercado del Trabajo y Salario Mínimo. Puntos de Referencia No. 93, Centro de Estudios Públicos. Card, D., & Krueger, A. B. (1995). Myth and measurement: the new economics of the minimum wage. Princeton University Press. De Loecker, J., Goldberg, P., Khandelwal, A., and N. Pavcnik (2016) Prices, Markups and Trade Reform Econometrica 84(2), De Fraja, G. (1999). Minimum wage legislation, productivity and employment. Economica, 66(264), Draca, M., Machin, S., & Van Reenen, J. (2011). Minimum wages and firm profitability. American economic journal: applied economics, 3(1), Eckel, C., & Neary, J. P. (2010). Multi-product firms and flexible manufacturing in the global economy. The Review of Economic Studies, 77(1), Goldberg, P. K., Khandelwal, A. K., Pavcnik, N., & Topalova, P. (2010). Multiproduct firms and product turnover in the developing world: Evidence from India. The Review of Economics and Statistics, 92(4), Hirsch, B. T., Kaufman, B. E., & Zelenska, T. (2015). Minimum wage channels of adjustment. Industrial Relations: A Journal of Economy and Society, 54(2),

23 Ma, Y., Tang, H., & Zhang, Y. (2014). Factor intensity, product switching, and productivity: Evidence from Chinese exporters. Journal of International Economics, 92(2), Machin, S., & Manning, A. (1996). Employment and the introduction of a minimum wage in Britain. The Economic Journal, 106(436), Mayer, T., Melitz, M. J., & Ottaviano, G. I. (2014). Market size, competition, and the product mix of exporters. The American Economic Review, 104(2), Mayneris, F., Poncet, S., & Zhang, T. (2014). The cleansing effect of minimum wage Minimum wage rules, firm dynamics and aggregate productivity in China. CEPII Working, No. Paper, 16. Nocke, V., & Yeaple, S. (2014). Globalization and multiproduct firms. International Economic Review, 55(4), Qiu, L. D., & Zhou, W. (2013). Multiproduct firms and scope adjustment in globalization. Journal of International Economics, 91(1), Schmitt, J. (2013). Why does the minimum wage have no discernible effect on employment? Center for Economic and Policy Research, 22, Wadsworth, J. (2010). Did the national minimum wage affect UK prices? Fiscal Studies, 31(1),

24 Figure 1 Real Minimum Wage Growth Rate: Source: Authors Elaboration based on Beyer (2008) 24

25 Figure 2 Product Level Exposure Distribution Density Exposure Observed US Predicted US 25

26 Figure 3 Previous Trends in Skilled and Unskilled Workers Unskilled Workers Skilled Workers (mean) logu (mean) logs year year Treated Control Treated Control 26

27 Figure 4 Previous Trends in Exporting and Licenses (mean) dexp Exporters % year (mean) dlic Foreign Licenses % year Treated Control Treated Control 27

28 Figure 5 Productivity Evolution and Product Creation and Destruction TFP levels year Both Add and Drop in at least 1 year in Others TFP difference difference year 28

29 Year Plants Products Table 1 Data Description: Plants and Products Products per Plant ISIC 6 digits ISIC 5 digits

30 Year Table 2 Average Product Destruction and Creation Rates by Year Drop Add Drop Add Total Sample Survivors(t,t+1) Average

31 Sector Table 3 Average Product Destruction and Creation Rates by Sector Drop Add Drop Add Total Sample Survivors (t,t+1) Food and Beverage Textile Wood Pulp and Paper Chemicals Non-metallic Metallic Machinery Other Industries Average

32 Table 4 Basic Model (1) (2) (3) (4) (5) (6) (7) (8) VARIABLES Drop Add Drop Add Drop Add Drop Add Q3EXP x MW 0.501*** *** 0.490*** *** [0.100] [0.063] [0.099] [0.063] ** ** [0.023] [0.016] [0.023] [0.016] *** *** *** *** [0.035] [0.024] [0.035] [0.024] ** *** ** *** [0.047] [0.032] [0.048] [0.032] Y/L *** *** *** *** [0.011] [0.007] [0.011] [0.007] EXP x MW 0.222*** ** 0.217*** ** [0.059] [0.037] [0.059] [0.037] Observations 36,511 33,230 36,511 33,230 36,511 33,230 36,511 33,230 R-squared Exp is the log of the unskill/skill wage bill ratio at the product level using product level data from single product firms. Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 32

33 Table 5 Surviving Plants (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW 0.311*** *** [0.074] [0.054] EXP x MW 0.109** ** [0.045] [0.031] Observations 30,713 28,327 30,713 28,327 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 33

34 Table 6 Single Product Plants (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3US x MW [0.308] [0.191] US x MW [0.141] [0.076] Observations 6,642 5,945 6,642 5,945 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 34

35 Table 7 Multi product plants (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW 0.485*** *** [0.107] [0.069] EXP x MW 0.221*** [0.064] [0.042] Observations 29,841 27,259 29,841 27,259 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 35

36 Table 8 Product Drop and Add in International Markets (1) (2) (3) (4) VARIABLES Drop X Add X Drop X Add X Q3EXP x MW [0.068] [0.039] EXP x MW [0.038] [0.028] Observations 36,511 33,230 36,511 33,230 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 36

37 Table 9 Heterogeneous Impact on Exported Products in Base year (1) (2) (3) (4) VARIABLES Drop Drop Drop Drop Q3EXP x MW 0.420*** 0.299*** [0.105] [0.077] Exported Product x MW *** ** *** *** [0.169] [0.162] [0.108] [0.096] Q3EXP x Exporter Product X MW 0.519* [0.271] [0.170] EXP x MW 0.200*** 0.119** [0.066] [0.048] EXP x Exporter Product X MW [0.121] [0.080] Observations 36,511 36,511 30,713 30,713 R-squared The estimation considers heterogeneous impacts on drop of products exported in 1996.Whole sample (columns 1-2), Survivors (columns 3-4). Robust standard errors clustered at the plantproduct level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 37

38 Table 10 Interactions with Sectorial GDP Growth (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW 0.409*** *** [0.102] [0.063] Q3EXP x dgdp Sector *** * [0.095] [0.075] EXP x MW 0.215*** * [0.061] [0.038] EXP x dgdp Sector ** [0.057] [0.043] Observations 36,511 33,230 36,511 33,230 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 38

39 Table 11 Interactions with Product Scope (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW 0.487*** *** [0.109] [0.060] Distance to scope xmw 0.063** *** 0.064** *** [0.026] [0.010] [0.026] [0.010] EXP x MW 0.229*** ** [0.066] [0.037] Observations 28,217 24,453 28,217 24,453 R-squared Distance to scope is the log difference in absolute value between the product code and the plant average product code. Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 39

40 Table 12 Impact on Core Products (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW 0.351*** *** [0.066] [0.038] EXP x MW 0.099*** *** [0.038] [0.022] Observations 36,511 33,230 36,511 33,230 R-squared A core product is a product with sales representing at least 75% of the plant total sales. Robust standard errors clustered at the plant-product level in brackets. All regressions include plantproduct and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 40

41 Table 13 Assigning EXP values randomly across products (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW [0.088] [0.054] EXP x MW [0.044] [0.026] Observations 33,272 30,550 33,272 30,550 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 41

42 Table 14 Exposure Variable Defined as Bite of minimum wage from CASEN 1996 (1) (2) (3) (4) (5) (6) (7) (8) VARIABLES Drop Add Drop Add Drop Add Drop Add Bite 1.2 x MW 0.700*** * 0.636*** *** [0.201] [0.111] [0.133] [0.080] Bite x MW 0.431** *** ** [0.175] [0.100] [0.106] [0.071] Observations 39,539 36,142 39,539 36,142 33,344 30,838 33,344 30,838 R-squared Bite is the fraction of workers earning up to 1.2 and 1minimum wages at the ISIC2-3digit level in logs, respectively. Whole sample (columns 1-4), Survivors (columns 5-8). Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industryyear fixed effects. *** p<0.01, ** p<0.05, * p<0.1 42

43 Table 15 Product Destruction and Creation Defined at the 6-digit ISIC Level (1) (2) (3) (4) VARIABLES Drop Add Drop Add Q3EXP x MW 0.245** ** [0.102] [0.065] EXP x MW 0.115** [0.057] [0.038] Observations 29,058 26,765 29,058 26,765 R-squared Robust standard errors clustered at the plant-product level in brackets. All regressions include plant-product and industry-year fixed effects. *** p<0.01, ** p<0.05, * p<0.1 43